Towards the Next-Generation of Fault-Tolerant Control Systems

小故障如果不被发现，可能会导致灾难性的故障，造成重大的财产、人员和环境破坏。在每一种类型的复杂系统中，都有无数这样的例子。即使故障不会导致灾难性故障，它也可能导致严重的性能降级和过早故障。因此，需要先进的控制系统来检测大类系统中的故障开始，并最终减缓故障的发展。控制系统需要能够识别或分类故障，同时仍保持可控性。由于系统的非线性、传感器噪声和有限的传感器数据，使得故障检测在实际中变得极其困难。虽然所有系统都需要故障检测，但由于我最近取得的进展，这项提议的结果将最初通过诊断燃料电池和垂直轴风力涡轮机来推动清洁能源的创新。终极计划愿景：我的目标是充分推进故障检测和容错控制的研究，使故障不会引发灾难。在这一发现基金的整体研究计划中，我的目标是培养2名博士和2名硕士研究生，并有以下目标：1.提高完全非线性系统中的观测器设计。2.通过观察者加强机器学习系统。3.开发使用间歇性低功率传感器的观察者。4.设计观察器，以检测和容忍系统中的重大变化。预计这笔赠款还将推动未来在燃料电池组控制和诊断方面的几项合作。预计对弹性自动化的需求将出现爆炸性增长，我的目标是将这一计划扩大到清洁能源以外的系统。因此，从长远来看，该计划的目标是：1.将故障检测和容错控制扩展到无人驾驶汽车、仓库机器人和自动送货无人机。2.通过与应用材料和光子控制的计划合作，将故障检测和容错控制扩展到微芯片制造系统。3.提供无与伦比的下一代传感和控制策略HQP培训。这项研究在控制系统方面的进展有望减少燃料电池和风力系统的性能退化，从而降低它们的生命周期系统成本，从而加快向清洁能源的过渡。由于这项研究也适用于自动驾驶系统，因此通过提高自动驾驶汽车的安全性和可靠性会产生长期的影响。这项研究的潜力是广阔的，将有助于提高加拿大在控制系统方面的专业水平，并使SFU和加拿大在系统可靠性方面处于领先地位。

Small faults when left undetected can result in catastrophic failure with significant property, personnel, and environmental damages. There are countless such examples in every type of complex system. Even when a fault does not cause a catastrophic failure, it can lead to severe performance degradation and premature failure. Hence there is a need for advanced control systems to detect the inception of faults in a broad class of systems and ultimately mitigate its progression. Control systems need to be able to identify or classify faults, while still maintaining controllability. Fault detection is made extremely difficult in practice, owing to system nonlinearity, sensor noise, and limited sensor data. While fault detection is needed by all systems, owing to my recent progress the results from this proposal will initially drive innovations in clean energy through the diagnosis of fuel cells and vertical axis wind turbines. Ultimate Program Vision: I aim to sufficiently advance the study of fault detection and fault-tolerant control so that faults do not trigger catastrophes. Within the overall research program of this Discovery grant, I aim to train 2 PhD and 2 MASc students and have the following objectives: 1.Advancing observer design in fully nonlinear systems. 2.Enhancing machine learning systems through observers. 3.Developing observers for utilizing intermittent low-power sensors. 4.Designing observers to detect and tolerate large changes in systems. The grant is also anticipated to drive several future collaboration on the control and diagnostics of fuel cell stacks. There is an anticipated explosion in the need for resilient automation and I aims to expand this program to systems outside clean energy. Thus, in the longer-term, this program aims to 1.Extend fault detection and fault tolerance control to driver-less automobiles, warehouse robots, and automatic deliver-drones. 2.Extend fault detection and fault tolerance control to micro-chip manufacturing systems through planned collaboration with Applied Materials and Photon control. 3.Provide unparalleled HQP training on the next generation of sensing and control strategies. The advances in control systems from this research are expected to reduce performance degradations in fuel cell and wind systems, consequently reducing their lifetime system costs and therefore hasten the transition to clean energy. As this research is also applicable to autonomous systems, there are long-lasting impacts through the improvement of safety and reliability in autonomous vehicles. The potential of this research is broad and will help to both raise the level of Canadian expertise in control systems and position SFU and Canada as a leader in system reliability.